面向单响应与多响应任务的基于模拟的证据积累建模：eam 包
主持人: Wanke Pan(潘晚坷)
分享嘉宾
朱广予
澳大利亚国立大学（ANU）心理学博士
国际数学心理学学会会员，国际判断与决策学会会员，澳大利亚数学心理学学会会员，澳大利亚实验心理学会会员。主要从事决策科学、心理测量学与认知建模研究，重点关注个体在不确定性情境下的决策机制，并结合因果发现与计算建模方法，构建可解释的个体与群体决策过程的数学模型。相关成果发表于 Behavior Research Methods 及 Personality and Social Psychology Bulletin。
杨光
悉尼大学（USYD）心理学博士候选人
美国视觉科学学会会员，澳大利亚数学心理学学会会员，澳大利亚实验心理学会会员，关注心理学与计算机领域的交叉方向，如视觉感知过程的计算建模及方法学研究。相关成果展示于美国视觉科学学会、澳大利亚实验心理学会、及亚太视觉会议。负责基于模拟的证据积累建模 R 语言开源包 eam package 的高性能实现。
报告简介
Evidence accumulation models are widely used to model choices and reaction times in cognitive tasks. Despite their success in binary and multi-alternative decision-making, their application to sequential multi-response tasks, such as free recall and verbal fluency, has remained limited. The primary challenge in modeling multiple responses within a single trial lies in the absence of tractable likelihood functions. To address this limitation, we introduce simulation-based inference into the evidence accumulation framework and develop an R package that integrates simulation engines for five representative models (the diffusion decision model, leaky competing accumulator model, linear ballistic accumulator model, racing diffusion model, and Lévy flight model), extending them to the multi-response contexts. Across three simulation studies, the package demonstrated robust parameter recovery across representative models, their multi-response extensions, and covariate-dependent variants. An empirical application to an aging-related free recall dataset further demonstrated the package’s ability to uncover computational mechanisms underlying age-related differences in memory retrieval. Taken together, this package provides a foundation for building customized evidence accumulation models and expands the range of cognitive processes that can be studied within the evidence accumulation framework.
报告时间
北京时间 [GMT+8] 2026年06月05日 20:00~21:00
会议信息： ZOOM 会议号：863 0404 9478
报告语言： 中文
主持人： Wanke Pan(潘晚坷)
参考文献
- Zhu, G., & Yang, G. (Preprint). Simulation-Based Evidence Accumulation Modeling for Single- and Multi-Response Tasks: The eam Package. https://osf.io/preprints/psyarxiv/dg7c2_v1
其他
- 录屏（本场不录屏）
- 幻灯片
所属小组
DDMJC 主页: https://ddmjournalclub.github.io
DDM Bilibili 主页: https://space.bilibili.com/3461571806235136